How do I use ModelCheckpoint with custom metrics in Keras?

Yes it is possible.

Define custom metrics as described in the documentation:

import keras.backend as K

def mean_pred(y_true, y_pred):
    return K.mean(y_pred)

model.compile(optimizer='rmsprop',
              loss='binary_crossentropy',
              metrics=['accuracy', mean_pred])

      

        
        
        
      

    

To check all available metrics:

print(model.metrics_names)
> ['loss', 'acc', 'mean_pred']

      

        
        
        
      

    

Pass the metric name ModelCheckpoint

through monitor

. If you want the metric to be calculated in validation, use a prefix val_

.

ModelCheckpoint(weights.{epoch:02d}-{val_mean_pred:.2f}.hdf5,
                monitor='val_mean_pred',
                save_best_only=True,
                save_weights_only=True,
                mode='max',
                period=1)

      

        
        
        
      

    

Do not use mode='auto'

for custom metrics. Understand why here .

Why am I answering my own question? Check this one out .